3 ms·

This is obviously just a bit of fun, but looking at the output examples, I think the OCR used may be flawed in that it's heavily optimised to produce characters

This is obviously just a bit of fun, but looking at the output examples, I think the OCR used may be flawed in that it's heavily optimised to produce characters used in natural language (alpha & spaces) rather than to produce e.g. operators common in code. The number of "special characters" output is extremely small. This will bias it heavily toward producing identifiers, which will be valid in most common programming languages.

thomasahle 5y ago
The examples at the bottom of the page are mostly special characters, like the valid program `;i;c;;#\\?z{;?;;fn':.;, `
- hcrisp 5y ago
  Reminds me of the joke: "Perl is the only programming language which looks the same before and after encryption."
  - gpderetta 5y ago
    please, perl ha nothing on q
    - maximus-decimus 5y ago
      q reads like prose compared to J.
Wowfunhappy 5y ago
I don’t think that’s a flaw! The study is “95% of paint splatters” not “95% of random strings of ASCII characters”. If common OCR programs are optimized to produce letters and spaces, fair game! (You would of course want a “proper” scientific study to be more rigorous about these things.) Edit: Also, the programs in the article appear to be mostly special characters?
- lucideer 5y ago
  > Edit: Also, the programs in the article appear to be mostly special characters? I assume you mean the 4 listed in the "Errata" section? The full list is here: https://www.mcmillen.dev/sigbovik/splatters.html https://www.mcmillen.dev/sigbovik/splatters.html
- Dylan16807 5y ago
  > I don’t think that’s a flaw! The study is “95% of paint splatters” not “95% of random strings of ASCII characters”. If common OCR programs are optimized to produce letters and spaces, fair game! But if someone said "95% of paint splatters OCR as Welsh words" and it turned out that's because the OCR is looking for words, it doesn't say much interesting about Welsh like the title implies. > Edit: Also, the programs in the article appear to be mostly special characters? Look at the whole gallery though. Most of the OCR results are just a string of letters or a number. Those being valid is very mundane. Edit: Actually if you want to make the analogy super precise, you could say "oh yeah those strings of letters are just proper nouns, which makes them valid Welsh, and of course numbers are valid Welsh"
  - Wowfunhappy 5y ago
    > But if someone said "95% of paint splatters OCR as Welsh words" and it turned out that's because the OCR is looking for words, it doesn't say much interesting about Welsh like the title implies. I think it depends on the claim, right? IMO, a more analogous statement might be "95% of paint splatters OCR as grammatically-correct Welsh sentences." Unless the software is using something like GPT-3, that finding would be notable regardless of software tuning. Even if it tells us more about Welsh grammar than it does paint splatters. > Look at the whole gallery though. Most of the OCR results are just a string of letters or a number. Those being valid is very mundane. Yes, thanks, my mistake!
    - Dylan16807 5y ago
      > IMO, a more analogous statement might be "95% of paint splatters OCR as grammatically-correct Welsh sentences." Unless the software is using something like GPT-3, that finding would be notable regardless of software tuning. Even if it tells us more about Welsh grammar than it does paint splatters. Well, I'd argue that "potato" is a grammatically correct sentence in most languages. (Imagine someone asked you what kind of soup you're holding.) And so is "7" and so is "Fififz". And that's literally what's happening with Perl with the vast majority of these. It's treating a single word/number as a statement. If you want to say something notable about grammar, you need to be generating sentences that are longer than a single noun.
jamal-kumar 5y ago
what got me is the actual unrecognized text on one of the images in the bottom corner
coldcode 5y ago
This makes me wonder how many of Van Gogh's works are valid Perl. Perhaps he invented the language...
dekhn 5y ago
yes. the paper is really just sampling the OCR software's underlying distributions, combined with perl's acceptance of a wide range of things that look like line noise.

thomasahle 5y ago

The examples at the bottom of the page are mostly special characters, like the valid program `;i;c;;#\\?z{;?;;fn':.;, `

hcrisp 5y ago
Reminds me of the joke: "Perl is the only programming language which looks the same before and after encryption."
- gpderetta 5y ago
  please, perl ha nothing on q
  - maximus-decimus 5y ago
    q reads like prose compared to J.

Wowfunhappy 5y ago

I don’t think that’s a flaw! The study is “95% of paint splatters” not “95% of random strings of ASCII characters”. If common OCR programs are optimized to produce letters and spaces, fair game! (You would of course want a “proper” scientific study to be more rigorous about these things.) Edit: Also, the programs in the article appear to be mostly special characters?

lucideer 5y ago
> Edit: Also, the programs in the article appear to be mostly special characters? I assume you mean the 4 listed in the "Errata" section? The full list is here: https://www.mcmillen.dev/sigbovik/splatters.html https://www.mcmillen.dev/sigbovik/splatters.html
Dylan16807 5y ago
> I don’t think that’s a flaw! The study is “95% of paint splatters” not “95% of random strings of ASCII characters”. If common OCR programs are optimized to produce letters and spaces, fair game! But if someone said "95% of paint splatters OCR as Welsh words" and it turned out that's because the OCR is looking for words, it doesn't say much interesting about Welsh like the title implies. > Edit: Also, the programs in the article appear to be mostly special characters? Look at the whole gallery though. Most of the OCR results are just a string of letters or a number. Those being valid is very mundane. Edit: Actually if you want to make the analogy super precise, you could say "oh yeah those strings of letters are just proper nouns, which makes them valid Welsh, and of course numbers are valid Welsh"
- Wowfunhappy 5y ago
  > But if someone said "95% of paint splatters OCR as Welsh words" and it turned out that's because the OCR is looking for words, it doesn't say much interesting about Welsh like the title implies. I think it depends on the claim, right? IMO, a more analogous statement might be "95% of paint splatters OCR as grammatically-correct Welsh sentences." Unless the software is using something like GPT-3, that finding would be notable regardless of software tuning. Even if it tells us more about Welsh grammar than it does paint splatters. > Look at the whole gallery though. Most of the OCR results are just a string of letters or a number. Those being valid is very mundane. Yes, thanks, my mistake!
  - Dylan16807 5y ago
    > IMO, a more analogous statement might be "95% of paint splatters OCR as grammatically-correct Welsh sentences." Unless the software is using something like GPT-3, that finding would be notable regardless of software tuning. Even if it tells us more about Welsh grammar than it does paint splatters. Well, I'd argue that "potato" is a grammatically correct sentence in most languages. (Imagine someone asked you what kind of soup you're holding.) And so is "7" and so is "Fififz". And that's literally what's happening with Perl with the vast majority of these. It's treating a single word/number as a statement. If you want to say something notable about grammar, you need to be generating sentences that are longer than a single noun.

jamal-kumar 5y ago

what got me is the actual unrecognized text on one of the images in the bottom corner

coldcode 5y ago

This makes me wonder how many of Van Gogh's works are valid Perl. Perhaps he invented the language...

dekhn 5y ago

yes. the paper is really just sampling the OCR software's underlying distributions, combined with perl's acceptance of a wide range of things that look like line noise.